NVIDIA Agentic AI Certification Notes

This repository contains my study notes for the NVIDIA NCP Agentic AI (NCP-AAI) certification exam. I put this together while preparing for the certification because I wanted a single place to organize everything I was learning across the different exam domains.

The certification covers quite a lot of ground. The biggest chunk of the material revolves around NVIDIA Triton Inference Server, which is NVIDIA's open source inference serving platform. I spent a good amount of time understanding how Triton handles model deployment, including its support for multiple frameworks like TensorFlow, PyTorch, and ONNX Runtime all within the same serving infrastructure. The model repository structure, configuration files, and dynamic batching were areas I kept coming back to because they show up in various ways throughout the exam objectives.

Another core area is agentic AI architectures. This includes understanding how autonomous agents are designed, how they reason and plan, and how tool use fits into the picture. I found this section particularly interesting because it connects nicely with the hands on work I had done in my Udacity Agentic AI Nanodegree. Topics like ReAct patterns, chain of thought reasoning, and multi agent orchestration are covered here from NVIDIA's perspective, which means understanding how these patterns work with NVIDIA's ecosystem of tools and APIs.

The notes also cover NVIDIA NIM (NVIDIA Inference Microservices) and how it simplifies deploying optimized AI models. NIM provides pre-built containers with optimized inference engines, which is a different approach compared to setting up Triton from scratch. Understanding when to use NIM versus a custom Triton deployment was one of those practical distinctions I made sure to nail down.

I organized the notes by exam domain so they map directly to the official exam guide. Each section includes the key concepts, important details worth memorizing, and areas where I noticed the material going deeper than surface level. The goal was to create something useful not just for passing the exam but also as a reference for working with these tools in real projects.

If you are preparing for the same certification, feel free to use these notes as a supplement to the official NVIDIA training materials. They reflect my own learning path, so some areas might be more detailed than others based on where I needed to spend extra time.